IoT Security Based on Machine Learning
Rama Krishna Vanakamamidi, L. Ramalingam, N. Abirami, S. Priyanka, C. Sasi Kumar, S. Murugan
Abstract
The protection of user privacy and mitigation of threats like spoofing, DoS, jamming, and eavesdropping are essential for the Internets of Things (IoT) to fulfill its promise of bringing improved and intelligent services to users via the integration of diverse devices into networks. Machine learning (ML) approaches, including supervised learning, unsupervised learning, and Reinforcement Learning (RL), are explored as potential solutions for securing the IoT. This article focuses on the use of ML in IoT authentications, access controls, secured offloading, and malware detection strategies to safeguard sensitive information. The challenges of incorporating these ML-based security strategies into real-world IoT systems are also discussed. A complete analysis of the threats facing IoT networks is provided, as the methods already in use to protect against them and the need for securing such networks. Then, highlight the areas where these security solutions are lacking and might benefit from ML and DL methods. Finally, go through in-depth the current ML and DL methods for dealing with various IoT network security issues. Several potential research possibilities for IoT securities based on ML and DL are also discussed.